This is a first pass at using BIEN data to explore relationships between plant traits and range sizes. Input datasets include range size, functional traits, and habit.

First, a quick look at the input data. After joining the three datasets and doing some cleanup and filtering (e.g. I am ignorning traits and habits with less than 1000 records) the dataset contains 25030 records covering 4666 species. It’s very sparse – the typical species has values for just 1-2 traits.

I’m looking at three variables from the range size dataset: sampleSize (number of occurrence points), clippedHullArea (convex hull around points with water clipped out; this is N/A for species with <3 points), and updArea (a combo of these other two, in which species with no hull area are assigned the area of one grid cell; this results in a bimodal distribution).

Traits ~ Habit

Correlations among predictors – how do functional traits vary by life form?

Range ~ Traits + Habit

These are the main results – relationships between traits and range size, in log-log space, with 2.5% of the data removed on each end of both dimensions to deal with outliers. Linear trends are fit for each habit (colored) and overall (black). We can look at these relationships on three levels:

Species-level

Raw values, one point per species. Log-log scale.

Family-level

One point per family, representing mean across species. Log-log scale.

Within-family

Each species is compared only to other species in its family. Trait and range values are log-transformed, and then converted to within-family standard z-scores.

Range ~ Habit

How does range size vary by life form alone?

Range ~ Distinctiveness

Beyond functional traits, phylogenetic structure is another possible predictor of range size. Here I’m using species per genus (etc) as a very crude proxy for evolutionary age or distinctiveness. The sample size is much larger here here as we can use all species, not just those with known traits.

Possible next steps